δυσί, δυοῖν, δυεῖν
(Phryn. Ecl. 180, Phryn. Ecl. 181)

D. General commentary

The Atticist lexicographers, including Phrynichus (A.1, A.2), Aelius Dionysius (cited by Eustathius, B.3), and the author of the Philetaerus (B.1), were interested in the oblique cases of the numeral δύο ‘two’, for which several variants existed. In Homer this numeral is indeclinable (see Chantraine 1958, 260): in addition to being the normal nominative-accusative form, δύο / δύω functions as a genitive in Il. 10.253 τῶν δύο μοιράων ‘of the two parts’ (a verse rejected by Zenodotus, Aristophanes, and Aristarchus, cf. schol. (Ariston.) Hom. Il. 10.253a (A)) and in Od. 10.515 δύω ποταμῶν ἐριδούπων ‘of the two roaring rivers’, and as a dative in Il. 13.407 δύω κανόνεσσ᾿ ἀραρυῖαν ‘fitted with two rods’. Indeed, while the majority of scholars assumes that the cardinal numerals from ‘one’ to ‘four’ were declinable in Proto-Indo-European (see e.g. Chantraine 1961, 147–8; Szemerényi 1996, 221–2; Fortson 2010, 145–6), others have argued that ‘two’ was originally indeclinable and that the individual languages adapted it to particular inflectional patterns, most frequently by giving it dual endings (see Sihler 1995, 407–8; Waanders 1992, 370–1). Cowgill (1985; followed by EDG s.v. δύο) argued that PIE possessed an originally indeclinable form *duu̯o (reflected in Greek δύο, as well as in Germanic, Armenian, and Albanian forms), which was subsequently recharacterised as *duu̯o-H, with an overt nom.-acc. dual ending, reflected in Homeric δύω, Sanskrit d(u)vā́(u), etc. Be that as it may, Post-Homeric Greek also introduced dual endings in the oblique cases: the genitive-dative form δυοῖν, first attested in Pindar (C.1), became the norm in Attic drama and prose (e.g. C.6). Modern scholars have observed that in classical Attic δυοῖν normally accompanies substantives in the dual, unless they are abstract nouns, whereas δύο may be used with plural as well as dual nouns (see Rutherford 1881, 290, with references to earlier scholarship).

It is now necessary to provide a brief sketch of the origin and distribution of the variants δυοῖν, δυεῖν, and δυσί(ν); for a more detailed discussion, see AGP vol. 1, 247, 251–3. The form δυοῖν evolved into the later Attic variant δυεῖν by a phonetic development, probably a dissimilation /oi/ > /ei/ next to the labial vowel /y/, perhaps stimulated by the incipient monophthongisation of /oi/ to /y/, as originally suggested by Schwyzer (1939, 195–6). However, the chronology of the monophthongisation of /oi/ remains controversial. Although an early date has been proposed by Teodorsson (1974, 202–5), according to whom the development had been completed in the more innovative phonology of Attic-speakers by 350 BCE, other scholars (e.g. Threatte 1980, 337–8) dated it to the Roman period, which would be incompatible with the attestations of δυεῖν (Horrocks 2010, 162–3 includes this change among developments of the spoken koine, but refrains from assigning it a precise date). A dissimilation may be attested in Boeotian, judging from the form δουῖν /dui:n/ in Corinn. fr. 654 PMG col. 3.15, which West (1970, 284) compared to Attic δυεῖν, noting, however, that Wilamowitz’s emendation δουῦν (the expected Boeotian spelling corresponding to Att. δυοῖν) may be correct. In Attic, the same development is also reflected in the forms οἴκει < οἴκοι ‘at home’ and πεῖος < ποῖος ‘which’. Cartlidge (2014, 168–9) has recently hypothesised that the developments seen in οἴκει and δυεῖν are unconnected: the former would due to the analogy of other locative adverbs such as ἐκεῖ ‘there’ (hence with the long close-mid vowel /e:/ normally spelled <ει>; Schwyzer [1939, 549] had already posited an old adverb *οἰκεῖ, which then shifted to οἴκει both under the influence of οἴκοι and in order to distinguish it from οἰκεῖ); the latter would reflect an actual phonetic dissimilation that produced a new diphthong /ei/. Be that as it may, δυεῖν, οἴκει, and πεῖος were already recognised as late Attic forms in ancient scholarship: according to Eustathius (B.4), Heraclides of Miletus regarded the forms with οι as recent and those with ει as old, but it is likely that Eustathius reversed Heraclides’ doctrine by transposing the words ὑστέρας and ἀρχαία (see Cohn 1884, 106). Indeed, δυεῖν is attested from the 4th century BCE onwards, beginning with Aristotle’s Constitution of the Athenians (3x, see C.7) and New Comedy (Men. Dysc. 327 = C.9, frr. 203.6, 453.1; Hegesipp. fr. 1.6), although it also appears as a textual variant in earlier authors (e.g. Thphr. Char. 2.3, but see Diggle 2004, 187).

The form δυεῖν was in its turn superseded in the dative by the variant δυσί(ν), which arose by analogy with the dative plural of athematic stems, and in particular with τρισί(ν) ‘three’. It is first attested in the corpus Hippocraticum, while in Attic the first certain occurrences of δυσί(ν) are found in the corpus Aristotelicum (see C.8) and in Theophrastus; a fragment of the late Middle Comedy poet Timocles (C.5) apparently provides another early Attic occurrence (see F.2). In Attic inscriptionsInscriptions, δυοῖν is the norm for both the genitive and the dative throughout the classical period, whereas δυεῖν first occurs around 330 BCE (see C.4, from an inventory in which δυεῖν occurs once and δυοῖν five times), and becomes usual by the beginning of the 3rd century. The last attestation of δυοῖν dates to 285/4 BCE, apart from a single Roman-period occurrence of δυοῖν ὑπάτων ‘of the two consuls’ (IG 2².4251/3.1, end of the 2nd century CE). Owing to the conservatism of official inscriptions, δυσί(ν) is attested epigraphically only from the very end of the 3rd century BCE and continues in use until late antiquity. During the Roman period, δύο is used as a genitive, but never as a dative (see Threatte 1996, 415–6).

Incidentally, the analogical developments of Ionic and Attic are not isolated: the Greek dialects display a great variety of analogical forms of the numeral ‘two’, collected in DGE s.v. δύο. Many of these follow the analogy of the o-stems, including the dat. plur. δυοῖσι(ν) (Hippon. fr. 92.6 West² = 95.6 Degani, Hdt. 1.32.28, 1.104.11) or δυοῖς (Hsch. δ 2497, Cretan inscriptions); the gen. plur. δυῶν, attested in Herodotus (12x) and in dialectal inscriptions from several regions (Ionia, Crete, Delphi, Locris), may, of course, be based either on the thematic or on the athematic inflection (see E. for the possibly independent creation of an analogical gen. δύων/δυῶν in Medieval Greek).

Early koine prose authors, such as Polybius, generally employ δυεῖν in the genitive and δυσί(ν) in the dative, in accordance with its dative ending (see de Foucault 1972, 66). By contrast, the lower koine of the Ptolemaic papyriPapyri (on which see Mayser, Gramm. vol. 1,2, 71–3), the Septuagint (see Helbing 1907, 53), and the New Testament (see Blass, Debrunner 1976, 49) follows the norm of late Attic inscriptions, with δύο used in the genitive as well as in the direct cases, and δυσί(ν) in the dative. Papyri from the Roman and early Byzantine period present a more complex picture: the genitive is normally δύο, sometimes δυεῖν, and rarely δυοῖν or δυῶν, while the dative is normally δυσί(ν), occasionally δύο. Gignac (1981, 187–9) interprets δυοῖν as an influence from classical Attic, δυεῖν from late Attic, and δυῶν from literary Ionic and Doric.

The Atticists’ opinions on the above forms vary, although they all agree that δυοῖν is more correct than δυεῖν, since it respects analogy (the oblique cases of the dual end in -οῖν) and is attested in classical authors such as Demosthenes (C.6). The later variant δυεῖν is likewise deemed acceptable, since it occurs in the text of other canonical authors such as Thucydides (C.3, but see F.1). For the same reason, Thomas Magister (B.5) and Aelius Dionysius (via Eustathius, B.3) also approve the use of δύο in the oblique cases, as found in Thucydides (C.2). The lexica agree in proscribing the newest variant δυσί(ν), with the partial exception of Thomas Magister, who apparently regards it as an admissible form by highlighting its presence in Thucydides, though he most probably relies on a corrupted text (see F.1). Phrynichus (A.2) states that δυεῖν, though acceptable, should be used only as a genitive, not as a dative. Since in Ecl. 180 (A.1) δυοῖν is given as the correct alternative to δυσί(ν), Phrynichus apparently prescribed a paradigm with nom.-acc. δύο, gen. δυοῖν or δυεῖν, dat. δυοῖν. Rutherford (1881, 289–90) judged this rule ‘quite erroneous’ and assumed that the entry was a later addition, comparing the case of Ecl. 174Phryn. Ecl. 174 (the second part of the latter entry is indeed considered spurious by Fischer). Indeed, as observed above, the restriction of δυεῖν to the genitive is typical of koine prose, where the dative is instead δυσί(ν), whereas classical authors who use δυοῖν in the dative also use it as a genitive. It is therefore possible that Phrynichus attempted to combine into a single rule the usages of different authors. The same prescription is found in a scholium to Euripides (B.6). It should also be noted that some erudite sources argued for other distinctions between δυοῖν and δυεῖν, attributing them either to grammatical gender (B.2) or to the presence of the article, as in the grammatical miscellany transmitted in cod. Par. suppl. gr. 70 (ff. 190v–194v) and edited by Bachmann (1828 vol. 2, 373.2–3): δυεῖν δίχα τοῦ ἄρθρου γενικῆς ἐπαγομένης [...]· τοῖν δυοῖν δέ, μετὰ τοῦ ἄρθρου, ‘δυεῖν without the article, when a genitive is introduced […], but τοῖν δυοῖν with the article’. Rather, koine authors such as Polybius, who employ δυεῖν, pair it with the genitive plural article τῶν, not with the dual article τοῖν).

The Atticising writers avoid δυσί(ν), with the exception of Dio Chrysostom, who uses it regularly in the dative (6x), while reserving δυοῖν (5x) for the genitive, and of isolated occurrences in Aelius Aristides (C.10, where it may be due to the parallelism with τρισίν) and Lucian (2x). As a rule, the Atticising writers prefer to use δυοῖν in the genitive and δύο in the dative (see Schmidt, Atticismus vol. 4, 586–7). In this case, therefore, their praxis seems to have followed the lexicographer’s rules quite closely.

E. Byzantine and Modern Greek commentary

The classical genitive and dative forms of δύο continued to be used in literary texts throughout the medieval period. Remarkably, even authors writing in the higher registerRegister often employed all three variants, usually reserving δυσί(ν) for the dative and δυοῖν or the rarer δυεῖν for the genitive. Although this is not the place for a detailed investigation of their distribution in Byzantine sources, the case of Photius may serve as an example: he regularly employs δυσί(ν) in the dative (66x), while in the genitive he shows a preference for δυοῖν (32x) over δυεῖν (13x). On the usage of the chronicles, which do not attest to δυεῖν at all, but sometimes employ δυσί(ν) in the dative, see Psaltis (1913, 192).

In lower-register and non-literary sources, on the other hand, the numeral ‘two’ – insofar as it is spelled out rather than written with the Greek numeral β΄ or the Arabic numeral 2 – appears as δύο/δυό (the latter variant resulting from synizesis), which may function as a genitive in addition to the nominative and accusative. The inflected genitive is δύων/δυῶν, formed analogically after genitives plural in -ων, with rarer or regionally restricted variants such as δυῶνε, δυονῶ(ν)(ε), ἐδυονῶ(ν), or (ο)ἰδυό (CGMEMG vol. 2, 1238–9; see also Kriaras, LME s.v. δύο). In Modern Greek, δύο [ˈðio]/δυο [ðjo] has again become indeclinable, as it may already have been in the language’s prehistory. Nevertheless, a learned genitive δυονῶ(ν)(ε) or δυῶ(ν)(ε) is occasionally attested.

F. Commentary on individual texts and occurrences

(1)    Thuc. 8.101.1 (C.3)

This passage of Thucydides is quoted both by Eustathius (B.3), as evidence for the use of δυεῖν, and by Thomas Magister (B.5), with a different text (δυσὶν ἡμέραις), as proof that δυσί(ν) was an admissible form. The MSS of Thucydides do in fact read δυσὶν ἡμέραις, which, however, is rightly regarded by modern editors as corrupt. Lobeck (1820, 211) was the first to argue that the numeral should be corrected to δυοῖν, which is the norm in Plato and the orators, whereas the late form δυσί(ν) frequently intruded into the manuscript tradition of classical authors. Van Herwerden (1869, 134) objected that Thucydides always uses δυοῖν with a noun in the dual, but δύο with the plural, and therefore suggested that either δύο ἡμέραις or, more likely, δυοῖν ἡμέραιν should be read: the latter reading was indeed accepted by subsequent editors down to Alberti (2000, 303). Tosi (1988, 183–4) points to this case as a typical example of the dangers of Thomas Magister’s anomalist approach, which treated textually corrupted forms as genuine variants. In fact, it cannot be excluded that this is one of those entries in Thomas’ lexicon that were expanded by other scholars through the addition of further loci classici (see entry Thomas Magister, ’Ονομάτων Ἀττικῶν ἐκλογή). The δυεῖν in Eustathius’ text must likewise be a variant reading, since this form would be anachronistic in Thucydides.

(2)    Timocl. fr. 16.4–7 (C.5)

The occurrence of δυσίν in this fragment of Timocles’ Icarian Satyrs, transmitted by Ath. 8.339d (and preserved only in cod. A), would constitute one of the earliest attestations of a form otherwise unattested before the 3rd century BCE, together with a passage in Aristotle’s Politics (C.8), where δυσί(ν) occurs twice alongside δυοῖν. Indeed, Timocles’ play, whose genre is uncertain (either comedy or satyr drama), has a terminus ante quem of 324 BCE (see Apostolakis 2019, 141), making it roughly contemporary with, or only slightly later than, Aristotle’s treatise. Given the unexpectedly early date, one might wonder whether δυσίν is in fact a corruption of an earlier δυοῖν, under the influence of τισί at the end of the following line, if not a copying error caused by the palaeographic similarity between omicron and sigma in both majuscule and minuscule script. However, it is also possible that δυσίν was chosen precisely because the context contains four other datives plural rather than dual. If δυσίν is the genuine reading, it would suggest that this variant was already emerging in colloquial Attic in the 4th century BCE, even though it did not find its way into other written sources (on this fragment, see AGP vol. 1, 252).